There are an increasing number of disabled people in the
world. These people face many problems going about their
day to day lives, in order to improve the day to day lives of these
people, it is important to give much attention to the research of
artificial lower and upper limb prostheses Conventionally,
different pattern recognition and learning networks must be
developed for EMG signals extracted from different people, but
an exceptional method for pattern classification utilizing EMG
signals from forearm muscles of the upper limb is introduced in
this paper. This method allows the use of one network for
different people without dropping the accuracy, overcoming the
problem of individual difference during EMG signal collection.
This can be achieved in 2 different ways. The first way, 6
different time domain feature extraction methods are combined
using a regular pattern attaining 22 new features which are
used with 6 different main classifiers with a total of 22 sub
classifiers. This is done to identify which classifier gives the
highest classification accuracy. In the second method,
combining the feature extraction method using the sequence
(X, XY, Y) provides high accuracy and makes it possible to
use one network for classifying different people hand gesture
without any drop in the accuracy.